Device type state estimation

Abstract

A method for type state estimation of a user equipment connected to a wireless communication network. The method comprises updating, recursively, of a type state estimate. The type state estimate is a probability for the user equipment to be a drone conditioned on obtained kinematic state estimate updates concerning the user equipment. The user equipment is assigned to be a drone as a response on the type state estimate exceeding a threshold.

The invention claimed is: 
     
       1. A method for type state estimation of a user equipment connected to a wireless communication network, wherein said method comprising:
 updating, recursively, a type state estimate, being a probability for said user equipment to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment; and 
 assigning said user equipment to be a drone as a response on said type state estimate exceeding a threshold; and 
 wherein an updated type state estimate is equal to a preceding type state estimate probability multiplied with a likelihood of said obtained kinematic state estimate updates and normalized, and wherein said likelihood is a probability of a smooth indicator function, representing a discrimination feature of the kinematic state estimate, where said likelihood is conditioned on the user equipment being a drone. 
 
     
     
       2. The method for type state estimation according to  claim 1 , wherein said updating recursively said type state estimate is performed according to:
     P ( D|z   t )∝ P ( I ( {circumflex over (x)},f )| D ) P ( D|z   t-1 ),
 
 where P(D|z t ) is said probability for said user equipment to be a drone conditioned on a present kinematic state estimate update, P(D|z t-1 ) is said probability for said user equipment to be a drone conditioned on a preceding kinematic state estimate update and P(I({circumflex over (x)}, f)|D) is said likelihood of said obtained kinematic state estimate update, I({circumflex over (x)}, f) is said smooth indicator function, {circumflex over (x)} is said kinematic state estimate and f is the feature for which the likelihood is evaluated, and where t is time. 
 
     
     
       3. The method for type state estimation according to  claim 1 , wherein a model with two type states is used, in which one type state being said probability for said user equipment to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment and other type state being a probability for said user equipment not to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment. 
     
     
       4. The method for type state estimation according to  claim 3 , wherein:
 an updated type state estimate of said state of said user equipment not being a drone is equal to a preceding type state estimate probability multiplied with a likelihood of said obtained kinematic state estimate updates and normalized, in which said likelihood is a probability of a smooth indicator function, representing a discrimination feature of the kinematic state estimate, where said likelihood is conditioned on the user equipment not being a drone; or 
 said updating recursively of said type state estimate is performed according to:
     P (¬ D|z   t )∝ P ( I ( {circumflex over (x)},f )|¬ D ) P (¬ D|z   t-1 ),
 
 
 where P(¬D|z t ) is said probability for said user equipment to not be a drone conditioned on a present kinematic state estimate update, P(¬D|z t-1 ) is said probability for said user equipment to not be a drone conditioned on a preceding kinematic state estimate update and P(I({circumflex over (x)}, f)|¬D) is said likelihood of said obtained kinematic state estimate update, I({circumflex over (x)}, f) is said smooth indicator function, {circumflex over (x)} is said kinematic state estimate and f is the feature for which the likelihood is evaluated, and where t is time. 
 
     
     
       5. The method for type state estimation according to  claim 1 , wherein:
 said discrimination feature of the kinematic state estimate being selected from a list of:
 an altitude above ground of the kinematic state estimate; 
 an altitude velocity of the kinematic state estimate; 
 a horizontal speed of the kinematic state estimate; 
 a horizontal position of the kinematic state estimate; and 
 a magnitude of an acceleration of the kinematic state estimate; 
 
 said discrimination feature of the kinematic state estimate being modelled by a Gaussian probability distribution function in the smooth indicator function; 
 said updating recursively said type state estimate is performed conditioned on at least one of a kinematic state estimate accuracy and kinematic mode probability; 
 propagating said type state estimate to a present time; 
 propagating said type state estimate to a present time comprises diffusion of type probabilities towards a constant probability vector; 
 propagating said type state estimate to a present time is performed according to:
     P ( t+T,D|z   t )= P   D +( P ( t,D|z   t )− P   D )α −αT ,
 
 
 where P(t+T,D|z t ) is said type state estimate of a present time, (P(t,D|z t ) is said type state estimate of a previous time, P D  is said constant probability vector and α is a predetermined propagation constant; 
 said kinematic state estimate updates comprises estimated positions and velocities, covariances therefore, and mode probability information; or 
 any combination thereof. 
 
     
     
       6. A network node for type state estimation of a user equipment connected to a wireless communication network to which said network node is connected, the network node comprising:
 a processor; and 
 a memory comprising a computer program which, when executed by the processor, causes the network node to:
 update recursively a type state estimate, being a probability for said user equipment to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment; 
 assign said user equipment to be a drone as a response on said type state estimate exceeding a threshold; and 
 wherein an updated type state estimate is equal to a preceding type state estimate probability multiplied with a likelihood of said obtained kinematic state estimate updates and normalized, and wherein said likelihood is a probability of a smooth indicator function, representing a discrimination feature of the kinematic state estimate, where said likelihood is conditioned on the user equipment being a drone. 
 
 
     
     
       7. The network node for type estimation according to  claim 6 , wherein said network node to update recursively said type state estimate according to:
     P ( D|z   t )∝ P ( I ( {circumflex over (x)},f )| D ) P ( D|z   t-1 ),
 
 where P(D|z t ) is said probability for said user equipment to be a drone conditioned on a present kinematic state estimate update, P(D|z t-1 ) is said probability for said user equipment to be a drone conditioned on a preceding kinematic state estimate update and P(I({circumflex over (x)}, f)|D) is said likelihood of said obtained kinematic state estimate update, I({circumflex over (x)}, f) is said smooth indicator function, {circumflex over (x)} is said kinematic state estimate and f is the feature for which the likelihood is evaluated, and where t is time. 
 
     
     
       8. The network node according to  claim 6 , wherein a model with two type states is used, in which one type state being said probability for said user equipment to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment and other type state being a probability for said user equipment not to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment. 
     
     
       9. The network node according to  claim 8 , wherein:
 an updated type state estimate of said state of said user equipment not being a drone is equal to a preceding type state estimate probability multiplied with a likelihood of said obtained kinematic state estimate updates and normalized, in which said likelihood is a probability of a smooth indicator function, representing a discrimination feature of the kinematic state estimate, where said likelihood is conditioned on the user equipment not being a drone; or 
 update recursively said type state estimate according to:
     P (¬ D|z   t )∝ P ( I ( {circumflex over (x)},f )|¬ D ) P (¬ D|z   t-1 ),
 
 
 where P(¬D|z t ) is said probability for said user equipment to not be a drone conditioned on a present kinematic state estimate update, P(¬D|z t-1 ) is said probability for said user equipment to not be a drone ( 11 ) conditioned on a preceding kinematic state estimate update and P(I({circumflex over (x)}, f)|¬D) is said likelihood of said obtained kinematic state estimate update, I({circumflex over (x)}, f) is said smooth indicator function, {circumflex over (x)} is said kinematic state estimate and f is the feature for which the likelihood is evaluated, and where t is time. 
 
     
     
       10. The network node according to  claim 6 , wherein:
 said discrimination feature of the kinematic state estimate being selected from a list of:
 an altitude above ground of the kinematic state estimate; 
 an altitude velocity of the kinematic state estimate; 
 a horizontal speed of the kinematic state estimate; 
 a horizontal position of the kinematic state estimate; 
 a magnitude of an acceleration of the kinematic state estimate; 
 
 said discrimination feature of the kinematic state estimate being modelled by a Gaussian probability distribution function in the smooth indicator function; 
 said update recursively said type state estimate conditioned on at least one of a kinematic state estimate accuracy and kinematic mode probability; 
 further to propagate said type state estimate to a present time; 
 further to propagate said type state estimate to a present time by diffusion of type probabilities towards a constant probability vector; 
 further to propagate said type state estimate to a present time according to:
     P ( t+T,D|z   t )= P   D +( P ( t,D|z   t )− P   D )α −αT ,
 
 
 where P(t+T,D|z t ) is said type state estimate of a present time, (P(t,D|z t ) is said type state estimate of a previous time, P D  is said constant probability vector and α is a predetermined propagation constant; 
 said kinematic state estimate updates comprises estimated positions and velocities, covariances therefore, and mode probability information; or 
 any combination thereof. 
 
     
     
       11. A non-transitory computer-readable storage medium comprising instructions which, when executed by at least one processor is are capable of causing a network node for type state estimation of a user equipment connected to a wireless network to perform operations comprising:
 updating, recursively, a type state estimate, being a probability for said user equipment to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment; 
 assigning said user equipment to be a drone as a response on said type state estimate exceeding a threshold; and 
 wherein an updated type state estimate is equal to a preceding type state estimate probability multiplied with a likelihood of said obtained kinematic state estimate updates and normalized, and wherein said likelihood is a probability of a smooth indicator function, representing a discrimination feature of the kinematic state estimate, where said likelihood is conditioned on the user equipment being a drone. 
 
     
     
       12. The non-transitory computer-readable storage medium according to  claim 11 , wherein the instructions cause said updating recursively said type state estimate to be performed according to:
     P ( D|z   t )∝ P ( I ( {circumflex over (x)},f )| D ) P ( D|z   t-1 ),
 
 where P(D|z t ) is said probability for said user equipment to be a drone conditioned on a present kinematic state estimate update, P(D|z t-1 ) is said probability for said user equipment to be a drone conditioned on a preceding kinematic state estimate update and P(I({circumflex over (x)}, f)|D) is said likelihood of said obtained kinematic state estimate update, I({circumflex over (x)}, f) is said smooth indicator function, {circumflex over (x)} is said kinematic state estimate and f is the feature for which the likelihood is evaluated, and where t is time. 
 
     
     
       13. The non-transitory computer-readable storage medium according to  claim 11 , wherein the instructions cause a model with two type states to be used, in which one type state being said probability for said user equipment to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment and other type state being a probability for said user equipment not to be a drone conditioned on obtained kinematic state estimate updates concerning said user equipment. 
     
     
       14. The non-transitory computer-readable storage medium according to  claim 13 , wherein the instructions cause performing of operations where:
 an updated type state estimate of said state of said user equipment not being a drone is equal to a preceding type state estimate probability multiplied with a likelihood of said obtained kinematic state estimate updates and normalized, in which said likelihood is a probability of a smooth indicator function, representing a discrimination feature of the kinematic state estimate, where said likelihood is conditioned on the user equipment not being a drone; or 
 said updating recursively of said type state estimate is performed according to:
     P (¬ D|z   t )∝ P ( I ( {circumflex over (x)},f )|¬ D ) P (¬ D|z   t-1 ),
 
 
 where P(¬D|z t ) is said probability for said user equipment to not be a drone conditioned on a present kinematic state estimate update, P(¬D|z t-1 ) is said probability for said user equipment to not be a drone ( 11 ) conditioned on a preceding kinematic state estimate update and P(I({circumflex over (x)}, f)|¬D) is said likelihood of said obtained kinematic state estimate update, I({circumflex over (x)}, f) is said smooth indicator function, {circumflex over (x)} is said kinematic state estimate and f is the feature for which the likelihood is evaluated, and where t is time. 
 
     
     
       15. The non-transitory computer-readable storage medium according to  claim 11 , wherein the instructions cause performing of operations where:
 said discrimination feature of the kinematic state estimate being selected from a list of:
 an altitude above ground of the kinematic state estimate; 
 an altitude velocity of the kinematic state estimate; 
 a horizontal speed of the kinematic state estimate; 
 a horizontal position of the kinematic state estimate; 
 a magnitude of an acceleration of the kinematic state estimate; 
 
 said discrimination feature of the kinematic state estimate being modelled by a Gaussian probability distribution function in the smooth indicator function; 
 said updating recursively said type state estimate is performed conditioned on at least one of a kinematic state estimate accuracy and kinematic mode probability; 
 propagating said type state estimate to a present time; 
 propagating said type state estimate to a present time comprises diffusion of type probabilities towards a constant probability vector; 
 propagating said type state estimate to a present time is performed according to:
     P ( t+T,D|z   t )= P   D +( P ( t,D|z   t )− P   D )α −αT ,
 
 
 where P(t+T,D|z t ) is said type state estimate of a present time, (P(t,D|z t ) is said type state estimate of a previous time, P D  is said constant probability vector and α is a predetermined propagation constant; 
 said kinematic state estimate updates comprises estimated positions and velocities, covariances therefore, and mode probability information; or 
 any combination thereof.